Development of a hybrid LSTM with chimp optimization algorithm for the pressure ventilator prediction

Area of Science:

• Biomedical Engineering
• Computational Medicine
• Artificial Intelligence in Healthcare

Background:

• Mechanical ventilation is critical for severe pulmonary conditions, especially during pandemics.
• Autonomous adaptation of ventilator parameters is essential for optimal patient treatment.
• Accurate prediction of ventilator pressure is needed for safe and effective respiratory support.

Purpose of the Study:

• To develop a predictive model for mechanical ventilator pressure.
• To enhance prediction accuracy using advanced computational techniques.
• To improve patient management through precise pressure forecasting.

Main Methods:

• Utilized a Long Short-Term Memory (LSTM) computational model for forecasting ventilator pressure.
• Integrated the Chimp Optimization method (ChoA) to optimize LSTM hyperparameters.
• Developed the LSTM-ChoA model for enhanced predictive accuracy.

Main Results:

• The LSTM-ChoA model significantly outperformed existing optimization algorithms (GWO, WOA, PSO) and regression models (KNN, RF, SVM).
• Achieved a substantial reduction in Mean Square Error (MSE): ~14.8% vs. GWO, ~60% vs. PSO and WOA.
• Analysis of Variance (ANOVA) confirmed statistical significance with a p-value of 0.000.

Conclusions:

• The LSTM-ChoA model provides a highly accurate and statistically significant method for predicting mechanical ventilator pressure.
• This predictive capability can lead to improved autonomous control of ventilators and enhanced patient outcomes.
• The study highlights the potential of combining LSTM with ChoA for complex physiological modeling and medical device optimization.